FIG. 1 illustrates a known spectrum of a DSS sequence, which consists in NRZI form of a repeated 24 bipolar-symbol pattern as follows:p24={+1+1+1+1+1+1+1+1+1+1+1+1−1−1−1−1−1−1−1−1−1−1−1−1}.
This data set separator sequence can thus be regarded as a periodic square wave s(t) with a period 24T, where T denotes the symbol duration. A Fourier transform of this square wave is given by the following equation (1):
                                          S            ⁡                          (              f              )                                =                                    ∑              n                        ⁢                                          S                n                            ⁢                              δ                ⁡                                  (                                      f                    -                                          n                                              24                        ⁢                        T                                                                              )                                                                    ,                            (        1        )            with
      S    n    ⁢      {                                                      sinc              ⁡                              (                                  n                  /                  2                                )                                                          n                                odd                                                0                                n                                even                              .      
Hence, S(f) represents a line spectrum that is nonzero at odd frequencies and decreases in magnitude as 1/f.
The data set separator sequence illustrated in FIG. 1 has been used to identify the cyclic dibit response of the magnetic recording channel. The identified dibit response not only allows one to compute an equalizer (typically a zero-forcing equalizer or a minimum mean-square error equalizer) for the magnetic recording channel, but also enables the use of read channels with noise-predictive maximum likelihood (NPML) detection. NPML detection is needed in high-performance/high-capacity tape systems, where it provides a better match of the target characteristic to the physical channel characteristic as compared with known characteristics such as partial-response class 4 (PR4), extended PR4 (EPR4), etc. The efficient identification of the channel dibit response based on the DSS sequence is addressed here.